fsck reports errors but does not fix them.

[map7]

When I do a fsck /var I get the following errors:

** /dev/da0s1d (NO WRITE)
** Last Mounted on /var
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
UNREF FILE I=188420  OWNER=mysql MODE=100600
SIZE=0 MTIME=Dec  2 11:31 2008 
CLEAR? no

UNREF FILE I=188447  OWNER=mysql MODE=100600
SIZE=0 MTIME=Dec  2 11:31 2008 
CLEAR? no

UNREF FILE I=188482  OWNER=mysql MODE=100600
SIZE=0 MTIME=Dec  2 11:31 2008 
CLEAR? no

UNREF FILE I=188483  OWNER=mysql MODE=100600
SIZE=0 MTIME=Dec  2 11:31 2008 
CLEAR? no

** Phase 5 - Check Cyl groups
37401 files, 231756 used, 1220491 free (2611 frags, 152235 blocks, 0.2% fragmentation)

To try and fix these errors I first go into single user mode with:
# init 1

Then when I recheck that label in single user mode no errors are reported?
# fsck /dev/da0s1d

** /dev/da0s1d (NO WRITE)
** Last Mounted on /var
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
** Phase 5 - Check Cyl groups
37401 files, 231756 used, 1220491 free (2611 frags, 152235 blocks, 0.2% fragmentation)

How do I fix the errors?

I've also tried booting frenzy but this does not boot on my computer, it crashes the kernel and reboots.

[J65nko]

Have you tried

fsck -y

?

[jggimi]

The problem is no write access to the partition.

[map7]

Yes I have tried with the -y partition.

When I went to single user mode I made sure that /var /tmp and /usr were unmounted, so they should have write access.

The problems just don't appear when I'm doing a scan as a single user.

[map7]

I forgot to mention that I'm running this on a 3ware 9000 series controller in a 4 disk hardware RAID-5 configuration.

FreeBSD just sees this as one drive though.

[map7]

I've disabled the 3ware monitoring tool as I've seen people having problems with that.

[map7]

i thought it might be mysql locking files so I stopped mysql and did a fsck of /var and got other problems

** /dev/da0s1d (NO WRITE)
** Last Mounted on /var
** Phase 1 - Check Blocks and Sizes
** Phase 2 - Check Pathnames
** Phase 3 - Check Connectivity
** Phase 4 - Check Reference Counts
UNREF FILE  I=329912  OWNER=mysql MODE=100660
SIZE=5 MTIME=Dec  2 15:45 2008 
RECONNECT? no


CLEAR? no

** Phase 5 - Check Cyl groups
FREE BLK COUNT(S) WRONG IN SUPERBLK
SALVAGE? no

SUMMARY INFORMATION BAD
SALVAGE? no

BLK(S) MISSING IN BIT MAPS
SALVAGE? no

37424 files, 231871 used, 1220376 free (2616 frags, 152220 blocks, 0.2% fragmentation)

[map7]

I'm running FreeBSD 7.0 by the way.

[map7]

I just tried this under a single drive, FreeBSD 7.0 on another machine and I get the same issues with fsck.

It reports errors when it's mounted and live, but if I go into single user mode and unmount then there are no errors. Is this normal?

This didn't used to happen in FreeBSD 6.2, now it does and I cannot run dump on a live system as it keeps hanging on the UNREF files.

[ephemera]

Originally Posted by map7

It reports errors when it's mounted and live, but if I go into single user mode and unmount then there are no errors. Is this normal?

Umm, you are not supposed to run fsck on a mounted file system.
Either unmount the FS or mount it R/O in case of root.

[jggimi]

The errors you keep seeing, and not reading, say NO WRITE. Why do you think that is?

[map7]

I understand that to fix problems with fsck you must first unmount that filesystem. If you look at post #4 in this thread, you will see I have tried running fsck in single user mode with the filesystems unmounted.

Are you guys saying that I should ignore the errors I get when running fsck on a mounted filesystem because fsck is not supposed to be run this way? I know fsck will not fix errors on a live system, but does it report invalid problems with the filesystems if they are mounted?

[phoenix]

Originally Posted by map7

I understand that to fix problems with fsck you must first unmount that filesystem. If you look at post #4 in this thread, you will see I have tried running fsck in single user mode with the filesystems unmounted.

But, all the bits you posted showing fsck running, show "NO WRITE ACCESS".

Are you guys saying that I should ignore the errors I get when running fsck on a mounted filesystem because fsck is not supposed to be run this way? I know fsck will not fix errors on a live system, but does it report invalid problems with the filesystems if they are mounted?

If softupdates are enabled on the filesystem, then yes, the results you get from running fsck on a mounted filesystem will be undefined. Softupdates changes the way files are deleted, and space reclaimed.

[map7]

pheonix,

thank you for understanding the mess I've posted. I must not of explained myself correctly and posted inaccurate examples. What I was trying to show was the different results I got when running 'fsck' on a live filesystem compared to an unmounted filesystem. I was unsure if I was supposed to be getting errors when scanning a live filesystem, even though that filesystem has reported error free when unmounted.

I have 'soft updates' enabled for the filesystems in question, so that would explain these errors, thanks again.